In [341]:
columns = list(pdata) # Histogram of first 14 columns
pdata[columns].hist(stacked=False, bins=100, figsize=(15,40), layout=(14,2));
A significant number of hotel bookings are called-off due to cancellations or no-shows. The typical reasons for cancellations include change of plans, scheduling conflicts, etc. This is often made easier by the option to do so free of charge or preferably at a low cost which is beneficial to hotel guests but it is a less desirable and possibly revenue-diminishing factor for hotels to deal with. Such losses are particularly high on last-minute cancellations.
The new technologies involving online booking channels have dramatically changed customers’ booking possibilities and behavior. This adds a further dimension to the challenge of how hotels handle cancellations, which are no longer limited to traditional booking and guest characteristics.
The cancellation of bookings impact a hotel on various fronts:
The increasing number of cancellations calls for a Machine Learning based solution that can help in predicting which booking is likely to be canceled. Star Hotels Group has a chain of hotels in Portugal, they are facing problems with the high number of booking cancellations and have reached out to your firm for data-driven solutions. You as a data scientist have to analyze the data provided to find which factors have a high influence on booking cancellations, build a predictive model that can predict which booking is going to be canceled in advance, and help in formulating profitable policies for cancellations and refunds.
The data contains the different attributes of customers' booking details. The detailed data dictionary is given below.
Data Dictionary
import pandas as pd
import numpy as np
import datetime
import warnings
warnings.filterwarnings("ignore")
# Library to split data
from sklearn.model_selection import train_test_split
# libaries to help with data visualization
import matplotlib.pyplot as plt
import seaborn as sns
# Removes the limit for the number of displayed columns
pd.set_option("display.max_columns", None)
# Sets the limit for the number of displayed rows
pd.set_option("display.max_rows", 200)
pd.set_option('mode.chained_assignment', None)
# To build model for prediction
import statsmodels.stats.api as sms
from statsmodels.stats.outliers_influence import variance_inflation_factor
import statsmodels.api as sm
from statsmodels.tools.tools import add_constant
from sklearn.linear_model import LogisticRegression
import warnings
warnings.filterwarnings('ignore')
# To get diferent metric scores
from sklearn.metrics import (
f1_score,
accuracy_score,
recall_score,
precision_score,
confusion_matrix,
roc_auc_score,
plot_confusion_matrix,
precision_recall_curve,
roc_curve,
make_scorer
)
# Removes the limit for the number of displayed columns
pd.set_option("display.max_columns", None)
# Sets the limit for the number of displayed rows
pd.set_option("display.max_rows", 200)
# To tune different models
from sklearn.model_selection import GridSearchCV
from sklearn.tree import DecisionTreeClassifier
from sklearn import tree
pdata = pd.read_csv("StarHotelsGroup.csv")
pdata.tail()
| no_of_adults | no_of_children | no_of_weekend_nights | no_of_week_nights | type_of_meal_plan | required_car_parking_space | room_type_reserved | lead_time | arrival_year | arrival_month | arrival_date | market_segment_type | repeated_guest | no_of_previous_cancellations | no_of_previous_bookings_not_canceled | avg_price_per_room | no_of_special_requests | booking_status | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 56921 | 2 | 1 | 0 | 1 | Meal Plan 2 | 0 | Room_Type 4 | 45 | 2019 | 6 | 15 | Online | 0 | 0 | 0 | 163.880 | 1 | Not_Canceled |
| 56922 | 2 | 0 | 1 | 1 | Meal Plan 1 | 0 | Room_Type 1 | 320 | 2019 | 5 | 15 | Offline | 0 | 0 | 0 | 90.000 | 1 | Canceled |
| 56923 | 2 | 0 | 0 | 3 | Not Selected | 0 | Room_Type 1 | 63 | 2018 | 4 | 21 | Online | 0 | 0 | 0 | 94.500 | 0 | Canceled |
| 56924 | 2 | 0 | 2 | 2 | Not Selected | 0 | Room_Type 1 | 6 | 2019 | 4 | 28 | Online | 0 | 0 | 0 | 162.500 | 2 | Not_Canceled |
| 56925 | 2 | 0 | 1 | 2 | Meal Plan 1 | 0 | Room_Type 1 | 207 | 2018 | 12 | 30 | Offline | 0 | 0 | 0 | 161.670 | 0 | Not_Canceled |
pdata.shape#we have 56926 rows and 18 columns
(56926, 18)
pd.set_option(
"display.float_format", lambda x: "%.3f" % x)
pdata.describe().T
| count | mean | std | min | 25% | 50% | 75% | max | |
|---|---|---|---|---|---|---|---|---|
| no_of_adults | 56926.000 | 1.876 | 0.519 | 0.000 | 2.000 | 2.000 | 2.000 | 4.000 |
| no_of_children | 56926.000 | 0.111 | 0.409 | 0.000 | 0.000 | 0.000 | 0.000 | 10.000 |
| no_of_weekend_nights | 56926.000 | 0.836 | 0.876 | 0.000 | 0.000 | 1.000 | 2.000 | 8.000 |
| no_of_week_nights | 56926.000 | 2.262 | 1.432 | 0.000 | 1.000 | 2.000 | 3.000 | 17.000 |
| required_car_parking_space | 56926.000 | 0.026 | 0.160 | 0.000 | 0.000 | 0.000 | 0.000 | 1.000 |
| lead_time | 56926.000 | 93.714 | 92.408 | 0.000 | 21.000 | 65.000 | 142.000 | 521.000 |
| arrival_year | 56926.000 | 2018.248 | 0.645 | 2017.000 | 2018.000 | 2018.000 | 2019.000 | 2019.000 |
| arrival_month | 56926.000 | 6.490 | 3.027 | 1.000 | 4.000 | 6.000 | 9.000 | 12.000 |
| arrival_date | 56926.000 | 15.636 | 8.719 | 1.000 | 8.000 | 16.000 | 23.000 | 31.000 |
| repeated_guest | 56926.000 | 0.025 | 0.155 | 0.000 | 0.000 | 0.000 | 0.000 | 1.000 |
| no_of_previous_cancellations | 56926.000 | 0.021 | 0.326 | 0.000 | 0.000 | 0.000 | 0.000 | 13.000 |
| no_of_previous_bookings_not_canceled | 56926.000 | 0.168 | 1.944 | 0.000 | 0.000 | 0.000 | 0.000 | 72.000 |
| avg_price_per_room | 56926.000 | 109.611 | 38.256 | 0.000 | 85.000 | 105.000 | 129.700 | 540.000 |
| no_of_special_requests | 56926.000 | 0.666 | 0.814 | 0.000 | 0.000 | 0.000 | 1.000 | 5.000 |
pdata.duplicated().sum()
#the number of duplicates is high
14350
pdata.drop_duplicates(keep=False,inplace=True)#deleting the duplicates
pdata.isnull().sum()#lets check if there is any missing values
no_of_adults 0 no_of_children 0 no_of_weekend_nights 0 no_of_week_nights 0 type_of_meal_plan 0 required_car_parking_space 0 room_type_reserved 0 lead_time 0 arrival_year 0 arrival_month 0 arrival_date 0 market_segment_type 0 repeated_guest 0 no_of_previous_cancellations 0 no_of_previous_bookings_not_canceled 0 avg_price_per_room 0 no_of_special_requests 0 booking_status 0 dtype: int64
pdata.info()#checking the data type
<class 'pandas.core.frame.DataFrame'> Int64Index: 37725 entries, 1 to 56924 Data columns (total 18 columns): # Column Non-Null Count Dtype --- ------ -------------- ----- 0 no_of_adults 37725 non-null int64 1 no_of_children 37725 non-null int64 2 no_of_weekend_nights 37725 non-null int64 3 no_of_week_nights 37725 non-null int64 4 type_of_meal_plan 37725 non-null object 5 required_car_parking_space 37725 non-null int64 6 room_type_reserved 37725 non-null object 7 lead_time 37725 non-null int64 8 arrival_year 37725 non-null int64 9 arrival_month 37725 non-null int64 10 arrival_date 37725 non-null int64 11 market_segment_type 37725 non-null object 12 repeated_guest 37725 non-null int64 13 no_of_previous_cancellations 37725 non-null int64 14 no_of_previous_bookings_not_canceled 37725 non-null int64 15 avg_price_per_room 37725 non-null float64 16 no_of_special_requests 37725 non-null int64 17 booking_status 37725 non-null object dtypes: float64(1), int64(13), object(4) memory usage: 5.5+ MB
(booking_status)(market_segment_type)(room_type_reserved)(type_of_meal_plan)(avg_price_per_room)Questions:
columns = list(pdata) # Histogram of first 14 columns
pdata[columns].hist(stacked=False, bins=100, figsize=(15,40), layout=(14,2));
pd.set_option(
"display.float_format", lambda x: "%.3f" % x)
pdata.describe().T
| count | mean | std | min | 25% | 50% | 75% | max | |
|---|---|---|---|---|---|---|---|---|
| no_of_adults | 37725.000 | 1.931 | 0.531 | 0.000 | 2.000 | 2.000 | 2.000 | 4.000 |
| no_of_children | 37725.000 | 0.155 | 0.479 | 0.000 | 0.000 | 0.000 | 0.000 | 10.000 |
| no_of_weekend_nights | 37725.000 | 0.908 | 0.891 | 0.000 | 0.000 | 1.000 | 2.000 | 8.000 |
| no_of_week_nights | 37725.000 | 2.343 | 1.545 | 0.000 | 1.000 | 2.000 | 3.000 | 17.000 |
| required_car_parking_space | 37725.000 | 0.038 | 0.191 | 0.000 | 0.000 | 0.000 | 0.000 | 1.000 |
| lead_time | 37725.000 | 73.301 | 73.280 | 0.000 | 14.000 | 50.000 | 112.000 | 518.000 |
| arrival_year | 37725.000 | 2018.308 | 0.621 | 2017.000 | 2018.000 | 2018.000 | 2019.000 | 2019.000 |
| arrival_month | 37725.000 | 6.324 | 3.054 | 1.000 | 4.000 | 6.000 | 8.000 | 12.000 |
| arrival_date | 37725.000 | 15.655 | 8.824 | 1.000 | 8.000 | 16.000 | 23.000 | 31.000 |
| repeated_guest | 37725.000 | 0.034 | 0.181 | 0.000 | 0.000 | 0.000 | 0.000 | 1.000 |
| no_of_previous_cancellations | 37725.000 | 0.028 | 0.374 | 0.000 | 0.000 | 0.000 | 0.000 | 11.000 |
| no_of_previous_bookings_not_canceled | 37725.000 | 0.249 | 2.378 | 0.000 | 0.000 | 0.000 | 0.000 | 72.000 |
| avg_price_per_room | 37725.000 | 113.583 | 41.762 | 0.000 | 85.850 | 108.000 | 137.670 | 540.000 |
| no_of_special_requests | 37725.000 | 0.785 | 0.839 | 0.000 | 0.000 | 1.000 | 1.000 | 5.000 |
sns.countplot(data = pdata, x='no_of_adults')
plt.show()
sns.countplot(data = pdata, x='no_of_children')
plt.show()
sns.countplot(data = pdata, x='no_of_week_nights')
plt.show()
A number of weeknights (Monday to Friday) the guest stayed or booked to stay at the hotel. as you can see, 2, 1 and then 3 weeknights are on the top
sns.countplot(data = pdata, x='no_of_weekend_nights')
plt.show()
A number of weekend nights (Saturday or Sunday) the guest stayed or booked to stay at the hotel. As you see, most of the guests tend to do not book or stay at a hotel at the weekends.
sns.countplot(data = pdata, x='type_of_meal_plan')
plt.show()
plot of type of meal plan booked by the customer shows us that most of them just booked for Breakfast (Meal plan 1)
pdata['type_of_meal_plan'].value_counts()
Meal Plan 1 28251 Not Selected 7839 Meal Plan 2 1627 Meal Plan 3 8 Name: type_of_meal_plan, dtype: int64
plt.figure(figsize=(10,6))
pdata['required_car_parking_space'].value_counts().plot(kind='pie');
Only a few customers require car parking space
chart=sns.countplot(data = pdata, x='room_type_reserved')
chart.set_xticklabels(
chart.get_xticklabels(),
rotation=90,
horizontalalignment='right',
fontweight='light');
bar plot of type of room reserved by the customer shows that room type 1 is the most popular
plt.figure(figsize=(10,6))
sns.distplot(pdata.lead_time,color="b",bins=20);
The number of days between the date of booking and the guest's arrival date is highly skewed to the right It also shows that guests tend to book hotel rooms mostly between 1 to 100 days earlier.
sns.countplot(data = pdata, x='arrival_year')
plt.show()
arrival date for 2018 is on the top
sns.countplot(data = pdata, x='arrival_month')
plt.show()
most of the reservations of hotel rooms are for the month of April to August. also first and the last two months have the lowest reservation.
plt.figure(figsize=(10,6))
sns.countplot(data = pdata, x='arrival_date')
plt.show()
it seems that the guests do not book rooms very much for last day of month.
plt.figure(figsize=(10,15))
pdata['market_segment_type'].value_counts().plot(kind='pie',autopct='%1.1f%%');
online market segment with 82% is on the top
after that offline with 12% has been located at the second place
plt.figure(figsize=(10,7))
ax = (pdata.repeated_guest.value_counts()/len(pdata)*100).sort_index().plot(kind="bar", rot=0)
plt.title(" Repeated Guest")
for p in ax.patches:
ax.annotate('{:.2f}%'.format(p.get_height()), (p.get_x()+0.15, p.get_height()+1))
only 3.41% of the guests are repeated guest
plt.figure(figsize=(10,7))
asd = (pdata.no_of_previous_cancellations.value_counts()/len(pdata)*100).sort_index().plot(kind="bar", rot=0)
plt.title(" Previous Cancellations")
for p in asd.patches:
asd.annotate('{:.2f}%'.format(p.get_height()), (p.get_x()+0.15, p.get_height()+1))
almost all of the guests (99%) have no previous bookings that were canceled.
plt.figure(figsize=(10,6))
axx=sns.distplot(pdata.avg_price_per_room,color="b",bins=20)
mean = pdata['avg_price_per_room'].mean()
kdeline = axx.lines[0]
xs = kdeline.get_xdata()
ys = kdeline.get_ydata()
height = np.interp(mean, xs, ys)
axx.vlines(mean, 0, height, color='crimson', ls=':')
plt.title("Avg Price Of Per Room");
avg_price_per_room has a distribution skewed to the right, I think it looks like a normal distribution. its average is around 109 Euros.
sns.countplot(data = pdata, x='no_of_special_requests')
plt.title(" no_of_special_requests")
plt.show()
pdata['no_of_special_requests'].value_counts()
0 16581 1 14117 2 5755 3 1120 4 138 5 14 Name: no_of_special_requests, dtype: int64
fig,ax = plt.subplots(figsize=(15,15))
sns.heatmap(pdata.corr(), annot=True, linewidths=0.05, fmt= '.2f',cmap="cividis")
plt.show()
it seems that arrival_year and arrival month have correlation
plt.figure(figsize=(9,6))
sns.scatterplot(x='booking_status', y='avg_price_per_room',data=pdata,palette= "Paired");
plt.figure(figsize=(9,6))
sns.barplot(data = pdata, x='booking_status',y='no_of_special_requests',hue=pdata['arrival_year']);
difference between booking status and the number of special requests in different years. Totally, in 2019 booking the rooms with no cancelation and more number of requests are higher than others. in 2017 booking the rooms and its cancelation (with high special request) are higher than others.
plt.figure(figsize=(9,6))
sns.barplot(data = pdata, x='booking_status',y='lead_time');
The more the lead time, the more cancelation it is.
plt.figure(figsize=(9,6))
sns.barplot(data = pdata, x='booking_status',y='no_of_children');
Guests who have more children did cancel their reservations more than others.
plt.figure(figsize=(9,6))
sns.barplot(data = pdata, x='booking_status',y='no_of_adults');
it seems that nomber of adult has no effect on cancelation
# I figured out that the type of columns that are related to date are integer. after trying to convert them and
# got some errors, it was found that we have February 29 in year 2018 which is not correct. here I fix the problem
pdata['arrival_date']=np.where((pdata['arrival_month']==2)& (pdata['arrival_date']==29),28,pdata['arrival_date'])
df_ = pdata[['arrival_year','arrival_month','arrival_date']].copy()
df_.columns = ['year', 'month' , 'day']
pdata['date']=pd.to_datetime(df_)
plt.figure(figsize=(10,8))
sns.lineplot(data=pdata, x="date", y="avg_price_per_room");
There is a increasing in price of per room between 2017 to 2019 with a mild slope
pdata['arrival_month'].value_counts()
8 4839 7 4268 5 3755 3 3688 4 3676 6 3600 10 2677 2 2655 9 2622 12 2093 1 1931 11 1921 Name: arrival_month, dtype: int64
plt.figure(figsize=(8,6))
pdata['market_segment_type'].value_counts().plot(kind='pie',autopct='%1.1f%%');
plt.figure(figsize=(9,6))
sns.barplot(data = pdata, x='market_segment_type',y='avg_price_per_room');
plt.figure(figsize=(6,5))
pdata['booking_status'].value_counts().plot(kind='pie',autopct='%1.1f%%');
ff=pdata[pdata['booking_status']=='Canceled']
#ff['repeated_guest'].value_counts()
plt.figure(figsize=(8,7))
ff['repeated_guest'].value_counts().plot(kind='pie',autopct='%1.1f%%');
ff=pdata[pdata['booking_status']=='Canceled']
ff.repeated_guest.value_counts(normalize=True)
0 0.999 1 0.001 Name: repeated_guest, dtype: float64
plt.figure(figsize=(9,6))
sns.barplot(data = pdata, x='booking_status',y='no_of_special_requests');
numeric_columns = pdata.select_dtypes(include=np.number).columns.tolist()
# let's plot the boxplots of all columns to check for outliers
plt.figure(figsize=(20, 30))
for i, variable in enumerate(numeric_columns):
plt.subplot(5, 4, i + 1)
plt.boxplot(pdata[variable], whis=1.5)
plt.tight_layout()
plt.title(variable)
plt.show()
def treat_outliers(pdata, col):
"""
treats outliers in a variable
col: str, name of the numerical variable
df: dataframe
col: name of the column
"""
Q1 = pdata[col].quantile(0.25) # 25th quantile
Q3 = pdata[col].quantile(0.75) # 75th quantile
IQR = Q3 - Q1
Lower_Whisker = Q1 - 1.5 * IQR
Upper_Whisker = Q3 + 1.5 * IQR
# all the values smaller than Lower_Whisker will be assigned the value of Lower_Whisker
# all the values greater than Upper_Whisker will be assigned the value of Upper_Whisker
pdata[col] = np.clip(pdata[col], Lower_Whisker, Upper_Whisker)
return pdata
def treat_outliers_all(pdata, col_list):
"""
treat outlier in all numerical variables
col_list: list of numerical variables
df: data frame
"""
for c in col_list:
if (c!='no_of_adults')& (c!='no_of_children')& (c!='required_car_parking_space')&(c!='required_car_parking_space')& (c!='repeated_guest')& (c!='no_of_previous_cancellations')&(c!='no_of_previous_bookings_not_canceled'):
pdata = treat_outliers(pdata, c)
return pdata
# treating the outliers
numerical_col = pdata.select_dtypes(include=np.number).columns.tolist()
pdata = treat_outliers_all(pdata, numerical_col)
# let's look at the boxplots to see if the outliers have been treated or not
plt.figure(figsize=(20, 30))
for i, variable in enumerate(numeric_columns):
plt.subplot(5, 4, i + 1)
plt.boxplot(pdata[variable], whis=1.5)
plt.tight_layout()
plt.title(variable)
plt.show()
pdata["booking_status"] = pdata["booking_status"].apply(lambda x: 1 if x == "Canceled" else 0)
#Create Dummy Variables for 3 columns
df= pd.get_dummies(pdata, columns=['type_of_meal_plan','room_type_reserved','market_segment_type'])
X = df.drop(["booking_status","market_segment_type_Aviation","room_type_reserved_Room_Type 7","type_of_meal_plan_Meal Plan 2"], axis=1)
Y = df["booking_status"]
X =pd.get_dummies(X, drop_first=True)
# Splitting data in train and test sets
X_train, X_test, y_train, y_test = train_test_split(
X, Y, test_size=0.30, random_state=1,stratify=df['booking_status'])
pd.set_option(
"display.float_format", lambda x: "%.3f" % x)
pdata.describe().T
| count | mean | std | min | 25% | 50% | 75% | max | |
|---|---|---|---|---|---|---|---|---|
| no_of_adults | 37725.000 | 1.931 | 0.531 | 0.000 | 2.000 | 2.000 | 2.000 | 4.000 |
| no_of_children | 37725.000 | 0.155 | 0.479 | 0.000 | 0.000 | 0.000 | 0.000 | 10.000 |
| no_of_weekend_nights | 37725.000 | 0.907 | 0.886 | 0.000 | 0.000 | 1.000 | 2.000 | 5.000 |
| no_of_week_nights | 37725.000 | 2.305 | 1.390 | 0.000 | 1.000 | 2.000 | 3.000 | 6.000 |
| required_car_parking_space | 37725.000 | 0.038 | 0.191 | 0.000 | 0.000 | 0.000 | 0.000 | 1.000 |
| lead_time | 37725.000 | 72.143 | 69.574 | 0.000 | 14.000 | 50.000 | 112.000 | 259.000 |
| arrival_year | 37725.000 | 2018.308 | 0.621 | 2017.000 | 2018.000 | 2018.000 | 2019.000 | 2019.000 |
| arrival_month | 37725.000 | 6.324 | 3.054 | 1.000 | 4.000 | 6.000 | 8.000 | 12.000 |
| arrival_date | 37725.000 | 15.655 | 8.824 | 1.000 | 8.000 | 16.000 | 23.000 | 31.000 |
| repeated_guest | 37725.000 | 0.034 | 0.181 | 0.000 | 0.000 | 0.000 | 0.000 | 1.000 |
| no_of_previous_cancellations | 37725.000 | 0.028 | 0.374 | 0.000 | 0.000 | 0.000 | 0.000 | 11.000 |
| no_of_previous_bookings_not_canceled | 37725.000 | 0.249 | 2.378 | 0.000 | 0.000 | 0.000 | 0.000 | 72.000 |
| avg_price_per_room | 37725.000 | 113.154 | 39.619 | 8.120 | 85.850 | 108.000 | 137.670 | 215.400 |
| no_of_special_requests | 37725.000 | 0.764 | 0.782 | 0.000 | 0.000 | 1.000 | 1.000 | 2.500 |
| booking_status | 37725.000 | 0.333 | 0.471 | 0.000 | 0.000 | 0.000 | 1.000 | 1.000 |
pdata.shape# 14350 rows have been deleted.
(37725, 18)
columns = list(pdata) # Histogram of first 14 columns
pdata[columns].hist(stacked=False, bins=100, figsize=(15,40), layout=(14,2));
sns.countplot(data = pdata, x='no_of_adults')
plt.show()
guests with a number of 2 are higher
sns.countplot(data = pdata, x='no_of_children')
plt.show()
there is no change here. still, booking with no child is on the top
sns.countplot(data = pdata, x='no_of_week_nights')
plt.show()
after dealing with outliers it is still 2, 1, and then 3 weeknights which are on the top.
sns.countplot(data = pdata, x='no_of_weekend_nights')
plt.show()
as you see, most of the guests tend to do not book or stay at a hotel at the weekends.
sns.countplot(data = pdata, x='type_of_meal_plan')
plt.show()
The plot of type of meal plan booked by the customer shows us that most of them still booked for Breakfast (Meal plan 1)
chart=sns.countplot(data = pdata, x='room_type_reserved')
chart.set_xticklabels(
chart.get_xticklabels(),
rotation=90,
horizontalalignment='right',
fontweight='light');
no changes in room_type_reserved
plt.figure(figsize=(10,6))
sns.distplot(pdata.lead_time,color="b",bins=20);
the skewness of column lead_time is still to the right.
plt.figure(figsize=(10,7))
ax = (pdata.repeated_guest.value_counts()/len(pdata)*100).sort_index().plot(kind="bar", rot=0)
plt.title(" Repeated Guest")
for p in ax.patches:
ax.annotate('{:.2f}%'.format(p.get_height()), (p.get_x()+0.15, p.get_height()+1))
there is no changes. Still 3.41% of the guests are repeated guest
plt.figure(figsize=(10,6))
axx=sns.distplot(pdata.avg_price_per_room,color="b",bins=20)
mean = pdata['avg_price_per_room'].mean()
kdeline = axx.lines[0]
xs = kdeline.get_xdata()
ys = kdeline.get_ydata()
height = np.interp(mean, xs, ys)
axx.vlines(mean, 0, height, color='crimson', ls=':')
plt.title("Avg Price Of Per Room");
after dealing with outliers its skewness is better and still its mean is around 113.154 Euro
sns.countplot(data = pdata, x='no_of_special_requests')
plt.title(" no_of_special_requests")
plt.show()
Most of the customers had no special request
fig,ax = plt.subplots(figsize=(15,15))
sns.heatmap(pdata.corr(), annot=True, linewidths=0.05, fmt= '.2f',cmap="cividis")
plt.show()
plt.figure(figsize=(9,6))
sns.scatterplot(x='booking_status', y='avg_price_per_room',data=pdata,palette= "Paired");
There is no differences between cancelation of rooms with different prices.
plt.figure(figsize=(9,6))
sns.barplot(data = pdata, x='booking_status',y='no_of_special_requests',hue=pdata['arrival_year']);
I can't see any differences here from before deleting the outliers.
plt.figure(figsize=(9,6))
sns.barplot(data = pdata, x='booking_status',y='lead_time');
The more the lead time, the more cancelation it is.
plt.figure(figsize=(10,8))
sns.lineplot(data=pdata, x="date", y="avg_price_per_room");
#What are the busiest months in the hotel? busiest months in the hotel are still August and July
pdata['arrival_month'].value_counts()
8 4839 7 4268 5 3755 3 3688 4 3676 6 3600 10 2677 2 2655 9 2622 12 2093 1 1931 11 1921 Name: arrival_month, dtype: int64
# Which market segment do most of the guests come from?
plt.figure(figsize=(8,6))
pdata['market_segment_type'].value_counts().plot(kind='pie',autopct='%1.1f%%');
it is still Online market segment which is on top
#Hotel rates are dynamic and change according to demand and customer demographics.What are the differences in room prices in different market segments?¶
plt.figure(figsize=(9,6))
sns.barplot(data = pdata, x='market_segment_type',y='avg_price_per_room');
The online and complementary market segments have the highest and lowest average price, respectively.
#What percentage of bookings are canceled?
plt.figure(figsize=(6,5))
pdata['booking_status'].value_counts().plot(kind='pie',autopct='%1.1f%%');
33.3% of bookings have been canceled
#Repeating guests are the guests who stay in the hotel often and are important to brand equity. What percentage of repeating guests cancel?
ff=pdata[pdata['booking_status']==1]
#ff['repeated_guest'].value_counts()
plt.figure(figsize=(8,7))
ff['repeated_guest'].value_counts().plot(kind='pie',autopct='%1.1f%%');
almost none of them cancel their reservations :(0.1%)
#Many guests have special requirements when booking a hotel room. Do these requirements affect booking cancellation?
plt.figure(figsize=(9,6))
sns.barplot(data = pdata, x='booking_status',y='no_of_special_requests');
It seems that guests who have high special requirements tend to do not to cancel their books.
vif_series = pd.Series(
[variance_inflation_factor(X_train.values, i) for i in range(X_train.shape[1])],
index=X_train.columns,
dtype=float,
)
print("Series before feature selection: \n\n{}\n".format(vif_series))
Series before feature selection: no_of_adults 20.830 no_of_children 2.321 no_of_weekend_nights 2.184 no_of_week_nights 4.241 required_car_parking_space 1.079 lead_time 2.440 arrival_year 444.217 arrival_month 5.506 arrival_date 4.171 repeated_guest 2.218 no_of_previous_cancellations 1.565 no_of_previous_bookings_not_canceled 1.948 avg_price_per_room 20.458 no_of_special_requests 2.162 type_of_meal_plan_Meal Plan 1 19.259 type_of_meal_plan_Meal Plan 3 1.033 type_of_meal_plan_Not Selected 6.646 room_type_reserved_Room_Type 1 98.447 room_type_reserved_Room_Type 2 3.412 room_type_reserved_Room_Type 3 1.028 room_type_reserved_Room_Type 4 32.606 room_type_reserved_Room_Type 5 4.015 room_type_reserved_Room_Type 6 6.175 market_segment_type_Complementary 3.831 market_segment_type_Corporate 10.588 market_segment_type_Offline 27.377 market_segment_type_Online 180.890 dtype: float64
arrival_year, market_segment_type_Online, room_type_reserved_Room_Type 1 and some other columns exhibit high multicollinearity.
I will Remove arrival_year first to see which variable has a significant impact on the model's performance.
X_train1 = X_train.drop("arrival_year", axis=1)
vif_series1 = pd.Series(
[variance_inflation_factor(X_train1.values, i) for i in range(X_train1.shape[1])],
index=X_train1.columns,
)
print("Series before feature selection: \n\n{}\n".format(vif_series1))
Series before feature selection: no_of_adults 20.677 no_of_children 2.291 no_of_weekend_nights 2.171 no_of_week_nights 4.183 required_car_parking_space 1.079 lead_time 2.439 arrival_month 5.492 arrival_date 4.135 repeated_guest 2.173 no_of_previous_cancellations 1.565 no_of_previous_bookings_not_canceled 1.946 avg_price_per_room 18.476 no_of_special_requests 2.160 type_of_meal_plan_Meal Plan 1 17.389 type_of_meal_plan_Meal Plan 3 1.027 type_of_meal_plan_Not Selected 6.082 room_type_reserved_Room_Type 1 60.164 room_type_reserved_Room_Type 2 2.482 room_type_reserved_Room_Type 3 1.017 room_type_reserved_Room_Type 4 20.876 room_type_reserved_Room_Type 5 2.967 room_type_reserved_Room_Type 6 4.755 market_segment_type_Complementary 2.191 market_segment_type_Corporate 6.152 market_segment_type_Offline 14.791 market_segment_type_Online 100.527 dtype: float64
Now dropping market_segment_type_Online, to see which variable has a significant impact on the model's performance.
X_train2= X_train1.drop("market_segment_type_Online", axis=1)
vif_series2 = pd.Series(
[variance_inflation_factor(X_train2.values, i) for i in range(X_train2.shape[1])],
index=X_train2.columns,
)
print("Series before feature selection: \n\n{}\n".format(vif_series2))
Series before feature selection: no_of_adults 19.341 no_of_children 2.251 no_of_weekend_nights 2.168 no_of_week_nights 4.146 required_car_parking_space 1.078 lead_time 2.433 arrival_month 5.482 arrival_date 4.116 repeated_guest 2.165 no_of_previous_cancellations 1.563 no_of_previous_bookings_not_canceled 1.946 avg_price_per_room 16.861 no_of_special_requests 2.153 type_of_meal_plan_Meal Plan 1 16.032 type_of_meal_plan_Meal Plan 3 1.022 type_of_meal_plan_Not Selected 5.570 room_type_reserved_Room_Type 1 27.962 room_type_reserved_Room_Type 2 1.651 room_type_reserved_Room_Type 3 1.009 room_type_reserved_Room_Type 4 11.621 room_type_reserved_Room_Type 5 2.131 room_type_reserved_Room_Type 6 3.577 market_segment_type_Complementary 1.298 market_segment_type_Corporate 1.828 market_segment_type_Offline 1.391 dtype: float64
Now it is room_type_reserved_Room_Type 1 which has high VIF values so, I remove it
X_train3= X_train2.drop("room_type_reserved_Room_Type 1", axis=1)
vif_series3 = pd.Series(
[variance_inflation_factor(X_train3.values, i) for i in range(X_train3.shape[1])],
index=X_train3.columns,
)
print("Series before feature selection: \n\n{}\n".format(vif_series3))
Series before feature selection: no_of_adults 18.137 no_of_children 2.224 no_of_weekend_nights 2.146 no_of_week_nights 4.020 required_car_parking_space 1.078 lead_time 2.416 arrival_month 5.364 arrival_date 3.981 repeated_guest 2.160 no_of_previous_cancellations 1.563 no_of_previous_bookings_not_canceled 1.946 avg_price_per_room 15.193 no_of_special_requests 2.148 type_of_meal_plan_Meal Plan 1 10.388 type_of_meal_plan_Meal Plan 3 1.022 type_of_meal_plan_Not Selected 3.509 room_type_reserved_Room_Type 2 1.114 room_type_reserved_Room_Type 3 1.001 room_type_reserved_Room_Type 4 1.775 room_type_reserved_Room_Type 5 1.130 room_type_reserved_Room_Type 6 1.996 market_segment_type_Complementary 1.282 market_segment_type_Corporate 1.785 market_segment_type_Offline 1.334 dtype: float64
Now it is no_of_adults 1 which has high VIF values(18.137)
X_train4= X_train3.drop("no_of_adults", axis=1)
vif_series4 = pd.Series(
[variance_inflation_factor(X_train4.values, i) for i in range(X_train4.shape[1])],
index=X_train4.columns,
)
print("Series before feature selection: \n\n{}\n".format(vif_series4))
Series before feature selection: no_of_children 2.175 no_of_weekend_nights 2.133 no_of_week_nights 3.962 required_car_parking_space 1.078 lead_time 2.333 arrival_month 5.364 arrival_date 3.965 repeated_guest 2.156 no_of_previous_cancellations 1.562 no_of_previous_bookings_not_canceled 1.945 avg_price_per_room 11.393 no_of_special_requests 2.131 type_of_meal_plan_Meal Plan 1 8.879 type_of_meal_plan_Meal Plan 3 1.021 type_of_meal_plan_Not Selected 2.921 room_type_reserved_Room_Type 2 1.111 room_type_reserved_Room_Type 3 1.001 room_type_reserved_Room_Type 4 1.729 room_type_reserved_Room_Type 5 1.129 room_type_reserved_Room_Type 6 1.995 market_segment_type_Complementary 1.259 market_segment_type_Corporate 1.773 market_segment_type_Offline 1.291 dtype: float64
avg_price_per_room has high VIF values(11.393)
X_train5= X_train4.drop("avg_price_per_room", axis=1)
vif_series5 = pd.Series(
[variance_inflation_factor(X_train5.values, i) for i in range(X_train5.shape[1])],
index=X_train5.columns,
)
print("Series before feature selection: \n\n{}\n".format(vif_series5))
Series before feature selection: no_of_children 2.105 no_of_weekend_nights 2.129 no_of_week_nights 3.916 required_car_parking_space 1.074 lead_time 2.333 arrival_month 4.914 arrival_date 3.826 repeated_guest 2.155 no_of_previous_cancellations 1.562 no_of_previous_bookings_not_canceled 1.945 no_of_special_requests 2.079 type_of_meal_plan_Meal Plan 1 7.078 type_of_meal_plan_Meal Plan 3 1.019 type_of_meal_plan_Not Selected 2.454 room_type_reserved_Room_Type 2 1.103 room_type_reserved_Room_Type 3 1.001 room_type_reserved_Room_Type 4 1.526 room_type_reserved_Room_Type 5 1.046 room_type_reserved_Room_Type 6 1.904 market_segment_type_Complementary 1.170 market_segment_type_Corporate 1.762 market_segment_type_Offline 1.265 dtype: float64
type_of_meal_plan_Meal Plan 1 has higher VIF values than 5 it is (7.078) so I will remove it.
X_train6= X_train5.drop("type_of_meal_plan_Meal Plan 1", axis=1)
vif_series6 = pd.Series(
[variance_inflation_factor(X_train6.values, i) for i in range(X_train6.shape[1])],
index=X_train6.columns,
)
print("Series before feature selection: \n\n{}\n".format(vif_series6))
Series before feature selection: no_of_children 2.102 no_of_weekend_nights 2.085 no_of_week_nights 3.598 required_car_parking_space 1.072 lead_time 2.325 arrival_month 4.116 arrival_date 3.292 repeated_guest 2.152 no_of_previous_cancellations 1.561 no_of_previous_bookings_not_canceled 1.943 no_of_special_requests 2.012 type_of_meal_plan_Meal Plan 3 1.018 type_of_meal_plan_Not Selected 1.399 room_type_reserved_Room_Type 2 1.100 room_type_reserved_Room_Type 3 1.001 room_type_reserved_Room_Type 4 1.483 room_type_reserved_Room_Type 5 1.044 room_type_reserved_Room_Type 6 1.902 market_segment_type_Complementary 1.163 market_segment_type_Corporate 1.724 market_segment_type_Offline 1.249 dtype: float64
X_test_1 = X_test[X_train6.columns]
# defining a function to compute different metrics to check performance of a classification model built using statsmodels
def model_performance_classification_statsmodels(
model, predictors, target, threshold=0.5
):
"""
Function to compute different metrics to check classification model performance
model: classifier
predictors: independent variables
target: dependent variable
threshold: threshold for classifying the observation as class 1
"""
# checking which probabilities are greater than threshold
pred_temp = model.predict(predictors) > threshold
# rounding off the above values to get classes
pred = np.round(pred_temp)
acc = accuracy_score(target, pred) # to compute Accuracy
recall = recall_score(target, pred) # to compute Recall
precision = precision_score(target, pred) # to compute Precision
f1 = f1_score(target, pred) # to compute F1-score
# creating a dataframe of metrics
df_perf = pd.DataFrame(
{"Accuracy": acc, "Recall": recall, "Precision": precision, "F1": f1,},
index=[0],
)
return df_perf
# defining a function to plot the confusion_matrix of a classification model
def confusion_matrix_statsmodels(model, predictors, target, threshold=0.5):
"""
To plot the confusion_matrix with percentages
model: classifier
predictors: independent variables
target: dependent variable
threshold: threshold for classifying the observation as class 1
"""
y_pred = model.predict(predictors) > threshold
cm = confusion_matrix(target, y_pred)
labels = np.asarray(
[
["{0:0.0f}".format(item) + "\n{0:.2%}".format(item / cm.flatten().sum())]
for item in cm.flatten()
]
).reshape(2, 2)
plt.figure(figsize=(6, 4))
sns.heatmap(cm, annot=labels, fmt="")
plt.ylabel("True label")
plt.xlabel("Predicted label")
# fitting logistic regression model
logit = sm.Logit(y_train, X_train6.astype(float))
lg1 = logit.fit(disp=False)
print(lg1.summary())
Logit Regression Results
==============================================================================
Dep. Variable: booking_status No. Observations: 26407
Model: Logit Df Residuals: 26386
Method: MLE Df Model: 20
Date: Thu, 16 Sep 2021 Pseudo R-squ.: 0.2949
Time: 10:14:50 Log-Likelihood: -11841.
converged: False LL-Null: -16794.
Covariance Type: nonrobust LLR p-value: 0.000
========================================================================================================
coef std err z P>|z| [0.025 0.975]
--------------------------------------------------------------------------------------------------------
no_of_children 0.2030 0.046 4.461 0.000 0.114 0.292
no_of_weekend_nights -0.0481 0.018 -2.673 0.008 -0.083 -0.013
no_of_week_nights -0.0290 0.011 -2.614 0.009 -0.051 -0.007
required_car_parking_space -1.4284 0.115 -12.455 0.000 -1.653 -1.204
lead_time 0.0161 0.000 59.219 0.000 0.016 0.017
arrival_month -0.0605 0.005 -12.629 0.000 -0.070 -0.051
arrival_date -0.0150 0.002 -9.349 0.000 -0.018 -0.012
repeated_guest 1.3281 1.32e+04 0.000 1.000 -2.58e+04 2.58e+04
no_of_previous_cancellations -2.2558 1.32e+04 -0.000 1.000 -2.58e+04 2.58e+04
no_of_previous_bookings_not_canceled -29.0037 1.83e+05 -0.000 1.000 -3.58e+05 3.58e+05
no_of_special_requests -1.3043 0.024 -53.861 0.000 -1.352 -1.257
type_of_meal_plan_Meal Plan 3 20.8028 6844.903 0.003 0.998 -1.34e+04 1.34e+04
type_of_meal_plan_Not Selected -0.0847 0.040 -2.101 0.036 -0.164 -0.006
room_type_reserved_Room_Type 2 -0.7092 0.129 -5.489 0.000 -0.962 -0.456
room_type_reserved_Room_Type 3 0.1696 1.416 0.120 0.905 -2.607 2.946
room_type_reserved_Room_Type 4 0.1628 0.040 4.083 0.000 0.085 0.241
room_type_reserved_Room_Type 5 0.5362 0.108 4.954 0.000 0.324 0.748
room_type_reserved_Room_Type 6 0.4908 0.110 4.481 0.000 0.276 0.705
market_segment_type_Complementary -38.3289 1.24e+04 -0.003 0.998 -2.44e+04 2.43e+04
market_segment_type_Corporate -1.3212 0.131 -10.115 0.000 -1.577 -1.065
market_segment_type_Offline -2.7702 0.070 -39.705 0.000 -2.907 -2.634
========================================================================================================
The variables which have a high p-value mean that they are not significant therefore we can drop the complete variables
# running a loop to drop variables with high p-value
# initial list of columns
cols = X_train6.columns.tolist()
# setting an initial max p-value
max_p_value = 1
while len(cols) > 0:
# defining the train set
X_train_aux = X_train6[cols]
# fitting the model
model = sm.Logit(y_train, X_train_aux).fit(disp=False)
# getting the p-values and the maximum p-value
p_values = model.pvalues
max_p_value = max(p_values)
# name of the variable with maximum p-value
feature_with_p_max = p_values.idxmax()
if max_p_value > 0.05:
cols.remove(feature_with_p_max)
else:
break
selected_features = cols
print(selected_features)
['no_of_children', 'no_of_weekend_nights', 'no_of_week_nights', 'required_car_parking_space', 'lead_time', 'arrival_month', 'arrival_date', 'no_of_previous_cancellations', 'no_of_special_requests', 'room_type_reserved_Room_Type 2', 'room_type_reserved_Room_Type 4', 'room_type_reserved_Room_Type 5', 'room_type_reserved_Room_Type 6', 'market_segment_type_Corporate', 'market_segment_type_Offline']
X_train7 = X_train6[selected_features]
logit2 = sm.Logit(y_train, X_train7.astype(float))
lg2 = logit2.fit(disp=False)
print(lg2.summary())
Logit Regression Results
==============================================================================
Dep. Variable: booking_status No. Observations: 26407
Model: Logit Df Residuals: 26392
Method: MLE Df Model: 14
Date: Thu, 16 Sep 2021 Pseudo R-squ.: 0.2867
Time: 10:14:53 Log-Likelihood: -11979.
converged: True LL-Null: -16794.
Covariance Type: nonrobust LLR p-value: 0.000
==================================================================================================
coef std err z P>|z| [0.025 0.975]
--------------------------------------------------------------------------------------------------
no_of_children 0.2152 0.045 4.777 0.000 0.127 0.303
no_of_weekend_nights -0.0460 0.018 -2.563 0.010 -0.081 -0.011
no_of_week_nights -0.0272 0.011 -2.474 0.013 -0.049 -0.006
required_car_parking_space -1.4403 0.114 -12.605 0.000 -1.664 -1.216
lead_time 0.0164 0.000 60.217 0.000 0.016 0.017
arrival_month -0.0682 0.005 -14.634 0.000 -0.077 -0.059
arrival_date -0.0164 0.002 -10.435 0.000 -0.019 -0.013
no_of_previous_cancellations -2.1425 0.645 -3.319 0.001 -3.408 -0.877
no_of_special_requests -1.3064 0.024 -54.192 0.000 -1.354 -1.259
room_type_reserved_Room_Type 2 -0.7152 0.128 -5.579 0.000 -0.966 -0.464
room_type_reserved_Room_Type 4 0.1938 0.038 5.124 0.000 0.120 0.268
room_type_reserved_Room_Type 5 0.5455 0.106 5.139 0.000 0.337 0.754
room_type_reserved_Room_Type 6 0.5085 0.109 4.671 0.000 0.295 0.722
market_segment_type_Corporate -1.6876 0.125 -13.544 0.000 -1.932 -1.443
market_segment_type_Offline -2.7412 0.069 -39.607 0.000 -2.877 -2.606
==================================================================================================
Now no feature has p-value greater than 0.05, so we'll consider the features in X_train7 as the final ones and lg2 as final model.
Coefficient of smarket_segment_type_Offline, required_car_parking_space,arrival_year, arrival_month, no_of_previous_cancellations, and no_of_special_requests are negative increase in these will lead to decrease in chances of a guest cancel the reservation.
Coefficient of no_of_weekend_nights, no_of_week_nights, lead_time, room_type_reserved_Room_Type 4 & 1,avg_price_per_room ,type_of_meal_plan_Not Selected , market_segment_type_Online, are positive an increase in these will lead to increase in chances of a guest cancel the reservation.
# converting coefficients to odds
odds = np.exp(lg2.params)
# finding the percentage change
perc_change_odds = (np.exp(lg2.params) - 1) * 100
# removing limit from number of columns to display
pd.set_option("display.max_columns", None)
# adding the odds to a dataframe
pd.DataFrame({"Odds": odds, "Change_odd%": perc_change_odds}, index=X_train7.columns).T
| no_of_children | no_of_weekend_nights | no_of_week_nights | required_car_parking_space | lead_time | arrival_month | arrival_date | no_of_previous_cancellations | no_of_special_requests | room_type_reserved_Room_Type 2 | room_type_reserved_Room_Type 4 | room_type_reserved_Room_Type 5 | room_type_reserved_Room_Type 6 | market_segment_type_Corporate | market_segment_type_Offline | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Odds | 1.240 | 0.955 | 0.973 | 0.237 | 1.017 | 0.934 | 0.984 | 0.117 | 0.271 | 0.489 | 1.214 | 1.726 | 1.663 | 0.185 | 0.064 |
| Change_odd% | 24.005 | -4.495 | -2.688 | -76.314 | 1.651 | -6.590 | -1.625 | -88.264 | -72.921 | -51.090 | 21.380 | 72.554 | 66.273 | -81.504 | -93.551 |
no_of_children: Holding all other features constant a 1 unit change in no_of_adults will increase the odds of a cancelation by 1.2 times or a 24.0% increase in odds of cancel the book.
no_of_weekend_nights: Holding all other features constant a 1 unit change in the no_of_weekend_nights will decrease the odds of a guest cancel the book by 0.95 times or a decrease of 4.49% decrease in odds of a guest cancel the book.
no_of_week_nights: Holding all other features constant a 1 unit change in no_of_week_nights will decrease the odds of a cancelation by 0.97 times or a 2.68% decrease in odds of cancel the book.
required_car_parking_space: Holding all other features constant a 1 unit change in required_car_parking_space will decrease the odds of a cancelation by 0.23 times or a 76.3% decrease in odds of cancel the book.
lead_time: Holding all other features constant a 1 unit change in lead_time will increase the odds of a cancelation by 0.93 times or a 6.5% increase in odds of cancel the book.
arrival_month: Holding all other features constant a 1 unit change in arrival_month will decrease the odds of a cancelation by 0.94 times or a 5.0% decrease in odds of cancel the book.
no_of_previous_cancellations: Holding all other features constant a 1 unit change in no_of_previous_cancellations will decrease the odds of a cancelation by 0.117 times or a 88.0% decrease in odds of cancel the book
Interpretation for other attributes can be done similarly.
confusion_matrix_statsmodels(lg2, X_train7, y_train)
The confusion matrix
True Positives (TP): we correctly predicted that they do NOT CANCEL 58.37%
True Negatives (TN): we correctly predicted that they do CANCEL 20.11%
False Positives (FP): we incorrectly predicted that they do CANCEL (a "Type I error") 2213 Falsely predict positive Type I error (8.38%)
False Negatives (FN): we incorrectly predicted that they don't CANCEL (a "Type II error") 3471 Falsely predict negative Type II error(13.14%)
log_reg_model_train_perf = model_performance_classification_statsmodels(
lg2, X_train7, y_train
)
print("Training performance:")
log_reg_model_train_perf
Training performance:
| Accuracy | Recall | Precision | F1 | |
|---|---|---|---|---|
| 0 | 0.785 | 0.605 | 0.706 | 0.651 |
logit_roc_auc_train = roc_auc_score(y_train, lg2.predict(X_train7))
fpr, tpr, thresholds = roc_curve(y_train, lg2.predict(X_train7))
plt.figure(figsize=(7, 5))
plt.plot(fpr, tpr, label="Logistic Regression (area = %0.2f)" % logit_roc_auc_train)
plt.plot([0, 1], [0, 1], "r--")
plt.xlim([0.0, 1.0])
plt.ylim([0.0, 1.05])
plt.xlabel("False Positive Rate")
plt.ylabel("True Positive Rate")
plt.title("Receiver operating characteristic")
plt.legend(loc="lower right")
plt.show()
Logistic Regression model is giving a good performance on training set.
Let's see if the f1 score can be improved further, by changing the model threshold using AUC-ROC Curve
# Optimal threshold as per AUC-ROC curve
# The optimal cut off would be where tpr is high and fpr is low
fpr, tpr, thresholds = roc_curve(y_train, lg2.predict(X_train7))
optimal_idx = np.argmax(tpr - fpr)
optimal_threshold_auc_roc = thresholds[optimal_idx]
print(optimal_threshold_auc_roc)
0.36960342850037675
# creating confusion matrix
confusion_matrix_statsmodels(
lg2, X_train7, y_train, threshold=optimal_threshold_auc_roc
)
# checking model performance for this model
log_reg_model_train_perf_threshold_auc_roc = model_performance_classification_statsmodels(
lg2, X_train7, y_train, threshold=optimal_threshold_auc_roc
)
print("Training performance:")
log_reg_model_train_perf_threshold_auc_roc
Training performance:
| Accuracy | Recall | Precision | F1 | |
|---|---|---|---|---|
| 0 | 0.763 | 0.751 | 0.619 | 0.679 |
Precision and Accuracy of model have decreased but the other metrics have increased. The model is still giving a good performance.
y_scores = lg2.predict(X_train7)
prec, rec, tre = precision_recall_curve(y_train, y_scores,)
def plot_prec_recall_vs_tresh(precisions, recalls, thresholds):
plt.plot(thresholds, precisions[:-1], "b--", label="precision")
plt.plot(thresholds, recalls[:-1], "g--", label="recall")
plt.xlabel("Threshold")
plt.legend(loc="upper left")
plt.ylim([0, 1])
plt.figure(figsize=(10, 7))
plot_prec_recall_vs_tresh(prec, rec, tre)
plt.show()
At the threshold of 0.51, we get balanced recall and precision.
# setting the threshold
optimal_threshold_curve = 0.51
# creating confusion matrix
confusion_matrix_statsmodels(lg2, X_train7, y_train, threshold=optimal_threshold_curve)
log_reg_model_train_perf_threshold_curve = model_performance_classification_statsmodels(
lg2, X_train7, y_train, threshold=optimal_threshold_curve
)
print("Training performance:")
log_reg_model_train_perf_threshold_curve
Training performance:
| Accuracy | Recall | Precision | F1 | |
|---|---|---|---|---|
| 0 | 0.785 | 0.595 | 0.712 | 0.649 |
Model is performing well on training set. There's not much improvement in the model performance as the default threshold is 0.36 and here we get 0.51 as the optimal threshold.
# training performance comparison
models_train_comp_df = pd.concat(
[
log_reg_model_train_perf.T,
log_reg_model_train_perf_threshold_auc_roc.T,
log_reg_model_train_perf_threshold_curve.T,
],
axis=1,
)
models_train_comp_df.columns = [
"Logistic Regression statsmodels",
"Logistic Regression 0.36 Threshold",
"Logistic Regression 0.51 Threshold",
]
print("Training performance comparison:")
models_train_comp_df
Training performance comparison:
| Logistic Regression statsmodels | Logistic Regression 0.36 Threshold | Logistic Regression 0.51 Threshold | |
|---|---|---|---|
| Accuracy | 0.785 | 0.763 | 0.785 |
| Recall | 0.605 | 0.751 | 0.595 |
| Precision | 0.706 | 0.619 | 0.712 |
| F1 | 0.651 | 0.679 | 0.649 |
All the logistic regression models have given a generalized performance on the training set. Recall score shows the best result on logestic regression 0.36 threshold (0.75).
X_test1 = X_test[list(X_train7.columns)]
# creating confusion matrix
confusion_matrix_statsmodels(lg2, X_test1, y_test)
The confusion matrix
True Positives (TP): we correctly predicted that they do NOT CANCEL 58.16%
True Negatives (TN): we correctly predicted that they do CANCEL 20.14%
False Positives (FP): we incorrectly predicted that they do CANCEL (a "Type I error") 937 Falsely predict positive Type I error (8.60%)
False Negatives (FN): we incorrectly predicted that they don't CANCEL (a "Type II error") 1484 Falsely predict negative Type II error(13.11%)
log_reg_model_test_perf = model_performance_classification_statsmodels(
lg2, X_test1, y_test
)
print("Test performance:")
log_reg_model_test_perf
Test performance:
| Accuracy | Recall | Precision | F1 | |
|---|---|---|---|---|
| 0 | 0.783 | 0.606 | 0.701 | 0.650 |
logit_roc_auc_train = roc_auc_score(y_test, lg2.predict(X_test1))
fpr, tpr, thresholds = roc_curve(y_test, lg2.predict(X_test1))
plt.figure(figsize=(7, 5))
plt.plot(fpr, tpr, label="Logistic Regression (area = %0.2f)" % logit_roc_auc_train)
plt.plot([0, 1], [0, 1], "r--")
plt.xlim([0.0, 1.0])
plt.ylim([0.0, 1.05])
plt.xlabel("False Positive Rate")
plt.ylabel("True Positive Rate")
plt.title("Receiver operating characteristic")
plt.legend(loc="lower right")
plt.show()
Logistic Regression model is giving a good performance on TEST set.
# creating confusion matrix
confusion_matrix_statsmodels(lg2, X_test1, y_test, threshold=optimal_threshold_auc_roc)
The confusion matrix
True Positives (TP): we correctly predicted that they do NOT CANCEL 51.01%
True Negatives (TN): we correctly predicted that they do CANCEL 24.82%
False Positives (FP): we incorrectly predicted that they do CANCEL (a "Type I error") 1782 Falsely predict positive Type I error (15.74%)
False Negatives (FN): we incorrectly predicted that they don't CANCEL (a "Type II error") 954 Falsely predict negative Type II error(8.43%)
# checking model performance for this model
log_reg_model_test_perf_threshold_auc_roc = model_performance_classification_statsmodels(
lg2, X_test1, y_test, threshold=optimal_threshold_auc_roc
)
print("Test performance:")
log_reg_model_test_perf_threshold_auc_roc
Test performance:
| Accuracy | Recall | Precision | F1 | |
|---|---|---|---|---|
| 0 | 0.758 | 0.746 | 0.612 | 0.672 |
# creating confusion matrix
confusion_matrix_statsmodels(lg2, X_test1, y_test, threshold=optimal_threshold_curve)
log_reg_model_test_perf_threshold_curve = model_performance_classification_statsmodels(
lg2, X_test1, y_test, threshold=optimal_threshold_curve
)
print("Test performance:")
log_reg_model_test_perf_threshold_curve
Test performance:
| Accuracy | Recall | Precision | F1 | |
|---|---|---|---|---|
| 0 | 0.784 | 0.594 | 0.709 | 0.646 |
Precision and Accuracy of model have increased but the other metrics have reduced. The model is still giving a good performance.
# testing performance comparison
models_test_comp_df = pd.concat(
[
log_reg_model_test_perf.T,
log_reg_model_test_perf_threshold_auc_roc.T,
log_reg_model_test_perf_threshold_curve.T,
],
axis=1,
)
models_test_comp_df.columns = [
"Logistic Regression statsmodels",
"Logistic Regression 0.36 Threshold",
"Logistic Regression 0.51 Threshold",
]
print("test performance comparison:")
models_test_comp_df
test performance comparison:
| Logistic Regression statsmodels | Logistic Regression 0.36 Threshold | Logistic Regression 0.51 Threshold | |
|---|---|---|---|
| Accuracy | 0.783 | 0.758 | 0.784 |
| Recall | 0.606 | 0.746 | 0.594 |
| Precision | 0.701 | 0.612 | 0.709 |
| F1 | 0.650 | 0.672 | 0.646 |
All the logistic regression models have given a generalized performance on the training and test set. Recall score shows the best result on logestic regression 0.36 threshold (0.74).
f1_score on the train and test sets are comparable.
## Function to calculate recall score
def get_recall_score(model, predictors, target):
"""
model: classifier
predictors: independent variables
target: dependent variable
"""
prediction = model.predict(predictors)
return recall_score(target, prediction)
def confusion_matrix_sklearn(model, predictors, target):
"""
To plot the confusion_matrix with percentages
model: classifier
predictors: independent variables
target: dependent variable
"""
y_pred = model.predict(predictors)
cm = confusion_matrix(target, y_pred)
labels = np.asarray(
[
["{0:0.0f}".format(item) + "\n{0:.2%}".format(item / cm.flatten().sum())]
for item in cm.flatten()
]
).reshape(2, 2)
plt.figure(figsize=(6, 4))
sns.heatmap(cm, annot=labels, fmt="")
plt.ylabel("True label")
plt.xlabel("Predicted label")
model = DecisionTreeClassifier(
criterion="gini", class_weight={0: 0.14, 1: 0.85}, random_state=1
)
model.fit(X_train7, y_train)
DecisionTreeClassifier(class_weight={0: 0.14, 1: 0.85}, random_state=1)
confusion_matrix_sklearn(model, X_train7, y_train)
decision_tree_perf_train = get_recall_score(model, X_train7, y_train)
print("Recall Score:", decision_tree_perf_train)
Recall Score: 1.0
confusion_matrix_sklearn(model, X_test1, y_test)
decision_tree_perf_test = get_recall_score(model, X_test1, y_test)
print("Recall Score:", decision_tree_perf_test)
Recall Score: 0.6221100186021791
## creating a list of column names
feature_names = X_train7.columns.to_list()
plt.figure(figsize=(20, 30))
out = tree.plot_tree(
model,
feature_names=feature_names,
filled=True,
fontsize=9,
node_ids=False,
class_names=None,
)
# below code will add arrows to the decision tree split if they are missing
for o in out:
arrow = o.arrow_patch
if arrow is not None:
arrow.set_edgecolor("black")
arrow.set_linewidth(1)
plt.show()
# importance of features in the tree building ( The importance of a feature is computed as the
# (normalized) total reduction of the criterion brought by that feature. It is also known as the Gini importance )
print(
pd.DataFrame(
model.feature_importances_, columns=["Imp"], index=X_train7.columns
).sort_values(by="Imp", ascending=False)
)
Imp lead_time 0.373305 arrival_date 0.147902 no_of_special_requests 0.118229 arrival_month 0.095939 no_of_week_nights 0.085332 market_segment_type_Offline 0.062460 no_of_weekend_nights 0.047092 room_type_reserved_Room_Type 4 0.017659 required_car_parking_space 0.016438 no_of_children 0.013268 market_segment_type_Corporate 0.009663 room_type_reserved_Room_Type 5 0.004255 room_type_reserved_Room_Type 2 0.003448 room_type_reserved_Room_Type 6 0.003265 no_of_previous_cancellations 0.001748
importances = model.feature_importances_
indices = np.argsort(importances)
plt.figure(figsize=(12, 12))
plt.title("Feature Importances")
plt.barh(range(len(indices)), importances[indices], color="violet", align="center")
plt.yticks(range(len(indices)), [feature_names[i] for i in indices])
plt.xlabel("Relative Importance")
plt.show()
According to the decision tree model, lead_time is the most important variable for predicting the cancelation.
# Choose the type of classifier.
estimator = DecisionTreeClassifier(random_state=1, class_weight={0: 0.14, 1: 0.85})
# Grid of parameters to choose from
parameters = {
"max_depth": [5, 10, 15, None],
"criterion": ["entropy", "gini"],
"splitter": ["best", "random"],
"min_impurity_decrease": [0.00001, 0.0001, 0.01],
}
# Type of scoring used to compare parameter combinations
scorer = make_scorer(recall_score)
# Run the grid search
grid_obj = GridSearchCV(estimator, parameters, scoring=scorer, cv=5)
grid_obj = grid_obj.fit(X_train7, y_train)
# Set the clf to the best combination of parameters
estimator = grid_obj.best_estimator_
# Fit the best algorithm to the data.
estimator.fit(X_train7, y_train)
DecisionTreeClassifier(class_weight={0: 0.14, 1: 0.85}, criterion='entropy',
max_depth=5, min_impurity_decrease=0.0001,
random_state=1, splitter='random')
confusion_matrix_sklearn(estimator, X_train7, y_train)
decision_tree_tune_perf_train = get_recall_score(estimator, X_train7, y_train)
print("Recall Score:", decision_tree_tune_perf_train)
Recall Score: 0.9711877918232548
confusion_matrix_sklearn(estimator, X_test1, y_test)
decision_tree_tune_perf_test = get_recall_score(estimator, X_test1, y_test)
print("Recall Score:", decision_tree_tune_perf_test)
Recall Score: 0.9697050225883603
plt.figure(figsize=(15, 10))
out = tree.plot_tree(
estimator,
feature_names=feature_names,
filled=True,
fontsize=9,
node_ids=False,
class_names=None,
)
for o in out:
arrow = o.arrow_patch
if arrow is not None:
arrow.set_edgecolor("black")
arrow.set_linewidth(1)
plt.show()
# Text report showing the rules of a decision tree -
print(tree.export_text(estimator, feature_names=feature_names, show_weights=True))
|--- lead_time <= 158.30 | |--- no_of_special_requests <= 1.62 | | |--- market_segment_type_Offline <= 0.79 | | | |--- no_of_special_requests <= 0.59 | | | | |--- lead_time <= 39.89 | | | | | |--- weights: [400.26, 1122.00] class: 1 | | | | |--- lead_time > 39.89 | | | | | |--- weights: [191.52, 2039.15] class: 1 | | | |--- no_of_special_requests > 0.59 | | | | |--- required_car_parking_space <= 0.54 | | | | | |--- weights: [828.94, 1431.40] class: 1 | | | | |--- required_car_parking_space > 0.54 | | | | | |--- weights: [55.44, 7.65] class: 0 | | |--- market_segment_type_Offline > 0.79 | | | |--- lead_time <= 55.72 | | | | |--- no_of_special_requests <= 0.60 | | | | | |--- weights: [164.36, 43.35] class: 0 | | | | |--- no_of_special_requests > 0.60 | | | | | |--- weights: [38.08, 0.00] class: 0 | | | |--- lead_time > 55.72 | | | | |--- lead_time <= 87.68 | | | | | |--- weights: [62.58, 22.10] class: 0 | | | | |--- lead_time > 87.68 | | | | | |--- weights: [79.10, 100.30] class: 1 | |--- no_of_special_requests > 1.62 | | |--- lead_time <= 122.32 | | | |--- no_of_special_requests <= 2.39 | | | | |--- no_of_weekend_nights <= 2.00 | | | | | |--- weights: [397.46, 141.95] class: 0 | | | | |--- no_of_weekend_nights > 2.00 | | | | | |--- weights: [1.26, 11.05] class: 1 | | | |--- no_of_special_requests > 2.39 | | | | |--- weights: [88.62, 0.00] class: 0 | | |--- lead_time > 122.32 | | | |--- no_of_special_requests <= 2.20 | | | | |--- lead_time <= 146.82 | | | | | |--- weights: [22.96, 52.70] class: 1 | | | | |--- lead_time > 146.82 | | | | | |--- weights: [5.74, 65.45] class: 1 | | | |--- no_of_special_requests > 2.20 | | | | |--- weights: [9.94, 0.00] class: 0 |--- lead_time > 158.30 | |--- market_segment_type_Offline <= 0.15 | | |--- no_of_special_requests <= 1.98 | | | |--- no_of_special_requests <= 0.76 | | | | |--- arrival_month <= 1.45 | | | | | |--- weights: [0.98, 16.15] class: 1 | | | | |--- arrival_month > 1.45 | | | | | |--- weights: [6.72, 1031.90] class: 1 | | | |--- no_of_special_requests > 0.76 | | | | |--- no_of_weekend_nights <= 0.67 | | | | | |--- weights: [6.02, 306.85] class: 1 | | | | |--- no_of_weekend_nights > 0.67 | | | | | |--- weights: [33.04, 490.45] class: 1 | | |--- no_of_special_requests > 1.98 | | | |--- no_of_special_requests <= 2.38 | | | | |--- room_type_reserved_Room_Type 4 <= 0.09 | | | | | |--- weights: [12.18, 330.65] class: 1 | | | | |--- room_type_reserved_Room_Type 4 > 0.09 | | | | | |--- weights: [0.84, 132.60] class: 1 | | | |--- no_of_special_requests > 2.38 | | | | |--- weights: [25.90, 0.00] class: 0 | |--- market_segment_type_Offline > 0.15 | | |--- weights: [35.70, 118.15] class: 1
# importance of features in the tree building ( The importance of a feature is computed as the
# (normalized) total reduction of the 'criterion' brought by that feature. It is also known as the Gini importance )
print(
pd.DataFrame(
estimator.feature_importances_, columns=["Imp"], index=X_train7.columns
).sort_values(by="Imp", ascending=False)
)
Imp lead_time 0.445663 no_of_special_requests 0.370906 market_segment_type_Offline 0.148377 required_car_parking_space 0.022753 no_of_weekend_nights 0.010187 room_type_reserved_Room_Type 4 0.001306 arrival_month 0.000807 no_of_children 0.000000 no_of_week_nights 0.000000 arrival_date 0.000000 no_of_previous_cancellations 0.000000 room_type_reserved_Room_Type 2 0.000000 room_type_reserved_Room_Type 5 0.000000 room_type_reserved_Room_Type 6 0.000000 market_segment_type_Corporate 0.000000
importances = estimator.feature_importances_
indices = np.argsort(importances)
plt.figure(figsize=(12, 12))
plt.title("Feature Importances")
plt.barh(range(len(indices)), importances[indices], color="violet", align="center")
plt.yticks(range(len(indices)), [feature_names[i] for i in indices])
plt.xlabel("Relative Importance")
plt.show()
In tuned decision tree lead_time is the most important feature followed by no_of_special_request.
clf = DecisionTreeClassifier(random_state=1, class_weight={0: 0.14, 1: 0.85})
path = clf.cost_complexity_pruning_path(X_train7, y_train)
ccp_alphas, impurities = path.ccp_alphas, path.impurities
pd.DataFrame(path)
| ccp_alphas | impurities | |
|---|---|---|
| 0 | 0.000 | 0.004 |
| 1 | 0.000 | 0.004 |
| 2 | 0.000 | 0.004 |
| 3 | 0.000 | 0.004 |
| 4 | 0.000 | 0.004 |
| ... | ... | ... |
| 2980 | 0.007 | 0.278 |
| 2981 | 0.007 | 0.285 |
| 2982 | 0.015 | 0.300 |
| 2983 | 0.018 | 0.336 |
| 2984 | 0.037 | 0.373 |
2985 rows × 2 columns
fig, ax = plt.subplots(figsize=(10, 5))
ax.plot(ccp_alphas[:-1], impurities[:-1], marker="o", drawstyle="steps-post")
ax.set_xlabel("effective alpha")
ax.set_ylabel("total impurity of leaves")
ax.set_title("Total Impurity vs effective alpha for training set")
plt.show()
lfs = []
for ccp_alpha in ccp_alphas:
clf = DecisionTreeClassifier(
random_state=1, ccp_alpha=ccp_alpha, class_weight={0: 0.14, 1: 0.85}
)
clf.fit(X_train7, y_train)
clfs.append(clf)
print(
"Number of nodes in the last tree is: {} with ccp_alpha: {}".format(
clfs[-1].tree_.node_count, ccp_alphas[-1]
)
)
Number of nodes in the last tree is: 1 with ccp_alpha: 0.037436862236564616
clfs = clfs[:-1]
ccp_alphas = ccp_alphas[:-1]
node_counts = [clf.tree_.node_count for clf in clfs]
depth = [clf.tree_.max_depth for clf in clfs]
fig, ax = plt.subplots(2, 1, figsize=(10, 7))
ax[0].plot(ccp_alphas, node_counts, marker="o", drawstyle="steps-post")
ax[0].set_xlabel("alpha")
ax[0].set_ylabel("number of nodes")
ax[0].set_title("Number of nodes vs alpha")
ax[1].plot(ccp_alphas, depth, marker="o", drawstyle="steps-post")
ax[1].set_xlabel("alpha")
ax[1].set_ylabel("depth of tree")
ax[1].set_title("Depth vs alpha")
fig.tight_layout()
recall_train = []
for clf in clfs:
pred_train = clf.predict(X_train7)
values_train = recall_score(y_train, pred_train)
recall_train.append(values_train)
recall_test = []
for clf in clfs:
pred_test = clf.predict(X_test1)
values_test = recall_score(y_test, pred_test)
recall_test.append(values_test)
train_scores = [clf.score(X_train7, y_train) for clf in clfs]
test_scores = [clf.score(X_test1, y_test) for clf in clfs]
fig, ax = plt.subplots(figsize=(15, 5))
ax.set_xlabel("alpha")
ax.set_ylabel("Recall")
ax.set_title("Recall vs alpha for training and testing sets")
ax.plot(
ccp_alphas, recall_train, marker="o", label="train", drawstyle="steps-post",
)
ax.plot(ccp_alphas, recall_test, marker="o", label="test", drawstyle="steps-post")
ax.legend()
plt.show()
# creating the model where we get highest train and test recall
index_best_model = np.argmax(recall_test)
best_model = clfs[index_best_model]
print(best_model)
DecisionTreeClassifier(ccp_alpha=0.00027836630003469305,
class_weight={0: 0.14, 1: 0.85}, random_state=1)
best_model.fit(X_train7, y_train)
DecisionTreeClassifier(ccp_alpha=0.00027836630003469305,
class_weight={0: 0.14, 1: 0.85}, random_state=1)
confusion_matrix_sklearn(best_model, X_train7, y_train)
print("Recall Score:", get_recall_score(best_model, X_train7, y_train))
Recall Score: 0.973920965721444
confusion_matrix_sklearn(best_model, X_test1, y_test)
print("Recall Score:", get_recall_score(best_model, X_test1, y_test))
Recall Score: 0.968110550093011
plt.figure(figsize=(15, 18))
out = tree.plot_tree(
best_model,
feature_names=feature_names,
filled=True,
fontsize=9,
node_ids=False,
class_names=None,
)
for o in out:
arrow = o.arrow_patch
if arrow is not None:
arrow.set_edgecolor("black")
arrow.set_linewidth(1)
plt.show()
best_model2 = DecisionTreeClassifier(
ccp_alpha=0.002, class_weight={0: 0.14, 1: 0.85}, random_state=1
)
best_model2.fit(X_train7, y_train)
DecisionTreeClassifier(ccp_alpha=0.002, class_weight={0: 0.14, 1: 0.85},
random_state=1)
confusion_matrix_sklearn(best_model2, X_train7, y_train)
decision_tree_postpruned_perf_train = get_recall_score(best_model2, X_train7, y_train)
print("Recall Score:", decision_tree_postpruned_perf_train)
Recall Score: 0.956838628857761
confusion_matrix_sklearn(best_model2, X_test1, y_test)
decision_tree_postpruned_perf_test = get_recall_score(best_model2, X_test1, y_test)
print("Recall Score:", decision_tree_postpruned_perf_test)
Recall Score: 0.9572149880414563
plt.figure(figsize=(15, 10))
out = tree.plot_tree(
best_model2,
feature_names=feature_names,
filled=True,
fontsize=9,
node_ids=False,
class_names=None,
)
for o in out:
arrow = o.arrow_patch
if arrow is not None:
arrow.set_edgecolor("black")
arrow.set_linewidth(1)
plt.show()
# training performance comparison
models_train_comp_df = pd.DataFrame(
[
decision_tree_perf_train,
decision_tree_tune_perf_train,
decision_tree_postpruned_perf_train,
],
columns=["Recall on training set"],
)
print("Training performance comparison:")
models_train_comp_df
Training performance comparison:
| Recall on training set | |
|---|---|
| 0 | 1.000000 |
| 1 | 0.971188 |
| 2 | 0.956839 |
# training performance comparison
models_test_comp_df = pd.DataFrame(
[
decision_tree_perf_test,
decision_tree_tune_perf_test,
decision_tree_postpruned_perf_test,
],
columns=["Recall on test set"],
)
print("testing performance comparison:")
models_test_comp_df
testing performance comparison:
| Recall on test set | |
|---|---|
| 0 | 0.622110 |
| 1 | 0.969705 |
| 2 | 0.957215 |
We analyzed the "StarHotels Project" using different techniques and used Decision Tree Classifier to build a predictive model for the same.
The model built can be used to predict if a customer is going to cancel their booking or not.
We visualized different trees and their confusion matrix to get a better understanding of the model. Easy interpretation is one of the key benefits of Decision Trees.
lead_time, no_of_special_requests and market_segment_type_Offlines are the most important variable in predicting the guests that will cancel the book. We established the importance of hyper-parameters/ pruning to reduce overfitting.
According to the decision tree model -
a) If a guest booking a room with lead_time less than or equal to 158.30 there's a very high chance the guest will not cancel his or her reservation.
b) If a guest booking a room with lead_time less than or equal to 158.30 and no_of_special_requests is less than or equal to 1.62 then there is a very high chance that the guest will not cancel his or her reservation.
c)If a guest booking a room with lead_time less than or equal to 158.30 and no_of_special_requests is less than or equal to 1.62 and market_segment_type_Offline is less than or equal 0.79 then there is a very high chance that the guest will not cancel his or her reservation.
It is observed that the less lead time in the reservation, less number of special requests, and fewer children lead to not cancelation guests booking. By present a good deal or service to people who have a high number of children or have a high special request it may lead to getting a lower chance of canceling between them.
It is observed that repeating guests, who stay in the hotel often and are important to brand equity will cancel their booking with a very small percentage (near to zero). so, I think offering a good deal for new guests for second booking can help the hotel to get a high number of repeat guests.
the online market segment with 82% is on the top of Market segment designation. I believe that managers should invest more in this segment.